In September 2026, an advanced AI model developed by OpenAI, called GPT-6 Astra, successfully deciphered a military code from World War I that had remained unsolved for 108 years. The message, sent by a German operator on November 27, 1918, contained details about the movements of British naval forces in the Crimea. The code used the ADFGVX cipher system, a complex encryption method used by Germany in the final months of the war. While hundreds of similar messages had been decrypted over the years—some by notable cryptanalysts like George Lasry—about a dozen remained unsolved, including this one, which was listed on a well-known list of 50 unsolved encrypted messages maintained by the German science portal Scienceblogs.de. The message was decrypted by a developer known only by the pseudonym Prinz, who submitted it to GPT-6 Astra, the latest and most advanced AI model from OpenAI, launched on September 3, 2026. The AI identified the encryption key "TRUPPENVERSCHIEBUNG," which means "movement of troops" in German, and produced the text: "A British cruiser has arrived at Sebastopol on the 24th. An Allied squadron follows on the 26th." The message, intended for a German supreme command, reported the movements of the Royal Navy in the Black Sea in the days following the armistice that ended World War I. The AI suggested that the message had resisted decryption for so long because the key "TRUPPENVERSCHIEBUNG" was not scheduled to be used until December 9, 1918—twelve days after the message was sent. The operator had used it prematurely, making it impossible for anyone relying on known keys and their documented usage dates to decrypt it. After producing the decryption, Astra automatically validated its results by retrieving original logbooks from the HMS Canterbury, a British light cruiser, confirming that the ship had indeed arrived at Sebastopol on November 24, 1918, and that an Allied squadron followed on November 26. The breakthrough sparked debate among cryptographic experts. Some suggested the task was not particularly difficult, as it involved using a known key outside its documented period of use. Others argued that the AI's ability to test unconventional hypotheses—ones that went against historical conventions—was what set it apart from traditional decryption methods. Astra seamlessly handled the decryption, hypothesis formation, and verification of historical records in a single, efficient process. This achievement followed OpenAI's earlier claims of solving the Navier-Stokes problem and decrypting WWII Enigma codes with the same model. In just three weeks, GPT-6 Astra achieved results that had long eluded human experts in fields ranging from cryptography to mathematics.